A long time ago, I made a tutorial on using the calibration points in Tracker Video. Of course many of my older posts are either lost or no longer correct. So, here is my new tutorial on calibration points.
What is a calibration point? First, think about a video. Or use this same one that I used in my tutorial - the 2012 USA Olympic Trials Men's 200m IM. There is some cool stuff going on here, but the camera doesn't full behave. After the start of the race, it both pans and zooms.
The calibration points are a way of tracking the motion of the camera. Essentially, you mark two points in the video that can be used to calibrate the coordinate system. By marking the location of these constant background points, Tracker video can move the origin and rescale the video based on the motion of these points.
Could you do this without the calibration points. However, that would mean for each frame you would have to both move the origin and adjust the scale. It is just much easier with the calibration points. Trust me on this one.
Ok, here is the tutorial.
After this, you can go ahead and mark all of your object points. A quick note. If the calibration points move off the screen, you can add another set. I would recommend adding the second set before the first set goes off the screen.
A Comment on Perspective
You know I can't just leave this alone, right? Just for fun, let me look at the problems with perspective in this video. From my best guess, the lanes in the pool have 1 meter sections of blue then white. Here is the lane line between lanes 4 and 5.
What does this mean? Well, I fit several points on a picture of the lanes. This means that the time variable isn't really time, it is just a number. For the portion of the data I fit a line to, it seems pretty linear. This means that the apparent length of each segment in the lane appears to be constant. What about the slope? The slope of -8.567 meters per second. But in this case, each successive point is counted as a step in time of 0.1 seconds (that was the default time step in Tracker). I could say this means 0.8567 meters per segment. That isn't what I expected. It's possible this is just a perspective error or my original calibration was off (likely).
What about the far lane (farthest from the camera)?
Again, it looks fairly linear but the data shows each segment to have an apparent length of 0.7208 meters. Since it is farther from the camera, it will have a smaller apparent size. Now for the closest lane.
Of course this shows a larger segment length at 1.14 meters. But what does this mean? It means that if you wanted to analyze the motion of all the swimmers, you would need to rescale the video for each swimmer. However, it also means that the data you get from Tracker for the one lane you are interested in will have a fairly constant scale as you move down the lane.
